contrib/diffutils/lib/localcharset.c

   1 /* Determine a canonical name for the current locale's character encoding.
   2
   3    Copyright (C) 2000-2006, 2008-2013 Free Software Foundation, Inc.
   4
   5    This program is free software; you can redistribute it and/or modify
   6    it under the terms of the GNU General Public License as published by
   7    the Free Software Foundation; either version 3, or (at your option)
   8    any later version.
   9
  10    This program is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13    GNU General Public License for more details.
  14
  15    You should have received a copy of the GNU General Public License along
  16    with this program; if not, see <http://www.gnu.org/licenses/>.  */
  17
  18 /* Written by Bruno Haible <bruno@clisp.org>.  */
  19
  20 #include <config.h>
  21
  22 /* Specification.  */
  23 #include "localcharset.h"
  24
  25 #include <fcntl.h>
  26 #include <stddef.h>
  27 #include <stdio.h>
  28 #include <string.h>
  29 #include <stdlib.h>
  30
  31 #if defined __APPLE__ && defined __MACH__ && HAVE_LANGINFO_CODESET
  32 # define DARWIN7 /* Darwin 7 or newer, i.e. Mac OS X 10.3 or newer */
  33 #endif
  34
  35 #if defined _WIN32 || defined __WIN32__
  36 # define WINDOWS_NATIVE
  37 #endif
  38
  39 #if defined __EMX__
  40 /* Assume EMX program runs on OS/2, even if compiled under DOS.  */
  41 # ifndef OS2
  42 #  define OS2
  43 # endif
  44 #endif
  45
  46 #if !defined WINDOWS_NATIVE
  47 # include <unistd.h>
  48 # if HAVE_LANGINFO_CODESET
  49 #  include <langinfo.h>
  50 # else
  51 #  if 0 /* see comment below */
  52 #   include <locale.h>
  53 #  endif
  54 # endif
  55 # ifdef __CYGWIN__
  56 #  define WIN32_LEAN_AND_MEAN
  57 #  include <windows.h>
  58 # endif
  59 #elif defined WINDOWS_NATIVE
  60 # define WIN32_LEAN_AND_MEAN
  61 # include <windows.h>
  62 #endif
  63 #if defined OS2
  64 # define INCL_DOS
  65 # include <os2.h>
  66 #endif
  67
  68 #if ENABLE_RELOCATABLE
  69 # include "relocatable.h"
  70 #else
  71 # define relocate(pathname) (pathname)
  72 #endif
  73
  74 /* Get LIBDIR.  */
  75 #ifndef LIBDIR
  76 # include "configmake.h"
  77 #endif
  78
  79 /* Define O_NOFOLLOW to 0 on platforms where it does not exist.  */
  80 #ifndef O_NOFOLLOW
  81 # define O_NOFOLLOW 0
  82 #endif
  83
  84 #if defined _WIN32 || defined __WIN32__ || defined __CYGWIN__ || defined __EMX__ || defined __DJGPP__
  85   /* Native Windows, Cygwin, OS/2, DOS */
  86 # define ISSLASH(C) ((C) == '/' || (C) == '\\')
  87 #endif
  88
  89 #ifndef DIRECTORY_SEPARATOR
  90 # define DIRECTORY_SEPARATOR '/'
  91 #endif
  92
  93 #ifndef ISSLASH
  94 # define ISSLASH(C) ((C) == DIRECTORY_SEPARATOR)
  95 #endif
  96
  97 #if HAVE_DECL_GETC_UNLOCKED
  98 # undef getc
  99 # define getc getc_unlocked
 100 #endif
 101
 102 /* The following static variable is declared 'volatile' to avoid a
 103    possible multithread problem in the function get_charset_aliases. If we
 104    are running in a threaded environment, and if two threads initialize
 105    'charset_aliases' simultaneously, both will produce the same value,
 106    and everything will be ok if the two assignments to 'charset_aliases'
 107    are atomic. But I don't know what will happen if the two assignments mix.  */
 108 #if __STDC__ != 1
 109 # define volatile /* empty */
 110 #endif
 111 /* Pointer to the contents of the charset.alias file, if it has already been
 112    read, else NULL.  Its format is:
 113    ALIAS_1 '\0' CANONICAL_1 '\0' ... ALIAS_n '\0' CANONICAL_n '\0' '\0'  */
 114 static const char * volatile charset_aliases;
 115
 116 /* Return a pointer to the contents of the charset.alias file.  */
 117 static const char *
 118 get_charset_aliases (void)
 119 {
 120   const char *cp;
 121
 122   cp = charset_aliases;
 123   if (cp == NULL)
 124     {
 125 #if !(defined DARWIN7 || defined VMS || defined WINDOWS_NATIVE || defined __CYGWIN__)
 126       const char *dir;
 127       const char *base = "charset.alias";
 128       char *file_name;
 129
 130       /* Make it possible to override the charset.alias location.  This is
 131          necessary for running the testsuite before "make install".  */
 132       dir = getenv ("CHARSETALIASDIR");
 133       if (dir == NULL || dir[0] == '\0')
 134         dir = relocate (LIBDIR);
 135
 136       /* Concatenate dir and base into freshly allocated file_name.  */
 137       {
 138         size_t dir_len = strlen (dir);
 139         size_t base_len = strlen (base);
 140         int add_slash = (dir_len > 0 && !ISSLASH (dir[dir_len - 1]));
 141         file_name = (char *) malloc (dir_len + add_slash + base_len + 1);
 142         if (file_name != NULL)
 143           {
 144             memcpy (file_name, dir, dir_len);
 145             if (add_slash)
 146               file_name[dir_len] = DIRECTORY_SEPARATOR;
 147             memcpy (file_name + dir_len + add_slash, base, base_len + 1);
 148           }
 149       }
 150
 151       if (file_name == NULL)
 152         /* Out of memory.  Treat the file as empty.  */
 153         cp = "";
 154       else
 155         {
 156           int fd;
 157
 158           /* Open the file.  Reject symbolic links on platforms that support
 159              O_NOFOLLOW.  This is a security feature.  Without it, an attacker
 160              could retrieve parts of the contents (namely, the tail of the
 161              first line that starts with "* ") of an arbitrary file by placing
 162              a symbolic link to that file under the name "charset.alias" in
 163              some writable directory and defining the environment variable
 164              CHARSETALIASDIR to point to that directory.  */
 165           fd = open (file_name,
 166                      O_RDONLY | (HAVE_WORKING_O_NOFOLLOW ? O_NOFOLLOW : 0));
 167           if (fd < 0)
 168             /* File not found.  Treat it as empty.  */
 169             cp = "";
 170           else
 171             {
 172               FILE *fp;
 173
 174               fp = fdopen (fd, "r");
 175               if (fp == NULL)
 176                 {
 177                   /* Out of memory.  Treat the file as empty.  */
 178                   close (fd);
 179                   cp = "";
 180                 }
 181               else
 182                 {
 183                   /* Parse the file's contents.  */
 184                   char *res_ptr = NULL;
 185                   size_t res_size = 0;
 186
 187                   for (;;)
 188                     {
 189                       int c;
 190                       char buf1[50+1];
 191                       char buf2[50+1];
 192                       size_t l1, l2;
 193                       char *old_res_ptr;
 194
 195                       c = getc (fp);
 196                       if (c == EOF)
 197                         break;
 198                       if (c == '\n' || c == ' ' || c == '\t')
 199                         continue;
 200                       if (c == '#')
 201                         {
 202                           /* Skip comment, to end of line.  */
 203                           do
 204                             c = getc (fp);
 205                           while (!(c == EOF || c == '\n'));
 206                           if (c == EOF)
 207                             break;
 208                           continue;
 209                         }
 210                       ungetc (c, fp);
 211                       if (fscanf (fp, "%50s %50s", buf1, buf2) < 2)
 212                         break;
 213                       l1 = strlen (buf1);
 214                       l2 = strlen (buf2);
 215                       old_res_ptr = res_ptr;
 216                       if (res_size == 0)
 217                         {
 218                           res_size = l1 + 1 + l2 + 1;
 219                           res_ptr = (char *) malloc (res_size + 1);
 220                         }
 221                       else
 222                         {
 223                           res_size += l1 + 1 + l2 + 1;
 224                           res_ptr = (char *) realloc (res_ptr, res_size + 1);
 225                         }
 226                       if (res_ptr == NULL)
 227                         {
 228                           /* Out of memory. */
 229                           res_size = 0;
 230                           free (old_res_ptr);
 231                           break;
 232                         }
 233                       strcpy (res_ptr + res_size - (l2 + 1) - (l1 + 1), buf1);
 234                       strcpy (res_ptr + res_size - (l2 + 1), buf2);
 235                     }
 236                   fclose (fp);
 237                   if (res_size == 0)
 238                     cp = "";
 239                   else
 240                     {
 241                       *(res_ptr + res_size) = '\0';
 242                       cp = res_ptr;
 243                     }
 244                 }
 245             }
 246
 247           free (file_name);
 248         }
 249
 250 #else
 251
 252 # if defined DARWIN7
 253       /* To avoid the trouble of installing a file that is shared by many
 254          GNU packages -- many packaging systems have problems with this --,
 255          simply inline the aliases here.  */
 256       cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
 257            "ISO8859-2" "\0" "ISO-8859-2" "\0"
 258            "ISO8859-4" "\0" "ISO-8859-4" "\0"
 259            "ISO8859-5" "\0" "ISO-8859-5" "\0"
 260            "ISO8859-7" "\0" "ISO-8859-7" "\0"
 261            "ISO8859-9" "\0" "ISO-8859-9" "\0"
 262            "ISO8859-13" "\0" "ISO-8859-13" "\0"
 263            "ISO8859-15" "\0" "ISO-8859-15" "\0"
 264            "KOI8-R" "\0" "KOI8-R" "\0"
 265            "KOI8-U" "\0" "KOI8-U" "\0"
 266            "CP866" "\0" "CP866" "\0"
 267            "CP949" "\0" "CP949" "\0"
 268            "CP1131" "\0" "CP1131" "\0"
 269            "CP1251" "\0" "CP1251" "\0"
 270            "eucCN" "\0" "GB2312" "\0"
 271            "GB2312" "\0" "GB2312" "\0"
 272            "eucJP" "\0" "EUC-JP" "\0"
 273            "eucKR" "\0" "EUC-KR" "\0"
 274            "Big5" "\0" "BIG5" "\0"
 275            "Big5HKSCS" "\0" "BIG5-HKSCS" "\0"
 276            "GBK" "\0" "GBK" "\0"
 277            "GB18030" "\0" "GB18030" "\0"
 278            "SJIS" "\0" "SHIFT_JIS" "\0"
 279            "ARMSCII-8" "\0" "ARMSCII-8" "\0"
 280            "PT154" "\0" "PT154" "\0"
 281          /*"ISCII-DEV" "\0" "?" "\0"*/
 282            "*" "\0" "UTF-8" "\0";
 283 # endif
 284
 285 # if defined VMS
 286       /* To avoid the troubles of an extra file charset.alias_vms in the
 287          sources of many GNU packages, simply inline the aliases here.  */
 288       /* The list of encodings is taken from the OpenVMS 7.3-1 documentation
 289          "Compaq C Run-Time Library Reference Manual for OpenVMS systems"
 290          section 10.7 "Handling Different Character Sets".  */
 291       cp = "ISO8859-1" "\0" "ISO-8859-1" "\0"
 292            "ISO8859-2" "\0" "ISO-8859-2" "\0"
 293            "ISO8859-5" "\0" "ISO-8859-5" "\0"
 294            "ISO8859-7" "\0" "ISO-8859-7" "\0"
 295            "ISO8859-8" "\0" "ISO-8859-8" "\0"
 296            "ISO8859-9" "\0" "ISO-8859-9" "\0"
 297            /* Japanese */
 298            "eucJP" "\0" "EUC-JP" "\0"
 299            "SJIS" "\0" "SHIFT_JIS" "\0"
 300            "DECKANJI" "\0" "DEC-KANJI" "\0"
 301            "SDECKANJI" "\0" "EUC-JP" "\0"
 302            /* Chinese */
 303            "eucTW" "\0" "EUC-TW" "\0"
 304            "DECHANYU" "\0" "DEC-HANYU" "\0"
 305            "DECHANZI" "\0" "GB2312" "\0"
 306            /* Korean */
 307            "DECKOREAN" "\0" "EUC-KR" "\0";
 308 # endif
 309
 310 # if defined WINDOWS_NATIVE || defined __CYGWIN__
 311       /* To avoid the troubles of installing a separate file in the same
 312          directory as the DLL and of retrieving the DLL's directory at
 313          runtime, simply inline the aliases here.  */
 314
 315       cp = "CP936" "\0" "GBK" "\0"
 316            "CP1361" "\0" "JOHAB" "\0"
 317            "CP20127" "\0" "ASCII" "\0"
 318            "CP20866" "\0" "KOI8-R" "\0"
 319            "CP20936" "\0" "GB2312" "\0"
 320            "CP21866" "\0" "KOI8-RU" "\0"
 321            "CP28591" "\0" "ISO-8859-1" "\0"
 322            "CP28592" "\0" "ISO-8859-2" "\0"
 323            "CP28593" "\0" "ISO-8859-3" "\0"
 324            "CP28594" "\0" "ISO-8859-4" "\0"
 325            "CP28595" "\0" "ISO-8859-5" "\0"
 326            "CP28596" "\0" "ISO-8859-6" "\0"
 327            "CP28597" "\0" "ISO-8859-7" "\0"
 328            "CP28598" "\0" "ISO-8859-8" "\0"
 329            "CP28599" "\0" "ISO-8859-9" "\0"
 330            "CP28605" "\0" "ISO-8859-15" "\0"
 331            "CP38598" "\0" "ISO-8859-8" "\0"
 332            "CP51932" "\0" "EUC-JP" "\0"
 333            "CP51936" "\0" "GB2312" "\0"
 334            "CP51949" "\0" "EUC-KR" "\0"
 335            "CP51950" "\0" "EUC-TW" "\0"
 336            "CP54936" "\0" "GB18030" "\0"
 337            "CP65001" "\0" "UTF-8" "\0";
 338 # endif
 339 #endif
 340
 341       charset_aliases = cp;
 342     }
 343
 344   return cp;
 345 }
 346
 347 /* Determine the current locale's character encoding, and canonicalize it
 348    into one of the canonical names listed in config.charset.
 349    The result must not be freed; it is statically allocated.
 350    If the canonical name cannot be determined, the result is a non-canonical
 351    name.  */
 352
 353 #ifdef STATIC
 354 STATIC
 355 #endif
 356 const char *
 357 locale_charset (void)
 358 {
 359   const char *codeset;
 360   const char *aliases;
 361
 362 #if !(defined WINDOWS_NATIVE || defined OS2)
 363
 364 # if HAVE_LANGINFO_CODESET
 365
 366   /* Most systems support nl_langinfo (CODESET) nowadays.  */
 367   codeset = nl_langinfo (CODESET);
 368
 369 #  ifdef __CYGWIN__
 370   /* Cygwin < 1.7 does not have locales.  nl_langinfo (CODESET) always
 371      returns "US-ASCII".  Return the suffix of the locale name from the
 372      environment variables (if present) or the codepage as a number.  */
 373   if (codeset != NULL && strcmp (codeset, "US-ASCII") == 0)
 374     {
 375       const char *locale;
 376       static char buf[2 + 10 + 1];
 377
 378       locale = getenv ("LC_ALL");
 379       if (locale == NULL || locale[0] == '\0')
 380         {
 381           locale = getenv ("LC_CTYPE");
 382           if (locale == NULL || locale[0] == '\0')
 383             locale = getenv ("LANG");
 384         }
 385       if (locale != NULL && locale[0] != '\0')
 386         {
 387           /* If the locale name contains an encoding after the dot, return
 388              it.  */
 389           const char *dot = strchr (locale, '.');
 390
 391           if (dot != NULL)
 392             {
 393               const char *modifier;
 394
 395               dot++;
 396               /* Look for the possible @... trailer and remove it, if any.  */
 397               modifier = strchr (dot, '@');
 398               if (modifier == NULL)
 399                 return dot;
 400               if (modifier - dot < sizeof (buf))
 401                 {
 402                   memcpy (buf, dot, modifier - dot);
 403                   buf [modifier - dot] = '\0';
 404                   return buf;
 405                 }
 406             }
 407         }
 408
 409       /* The Windows API has a function returning the locale's codepage as a
 410          number: GetACP().  This encoding is used by Cygwin, unless the user
 411          has set the environment variable CYGWIN=codepage:oem (which very few
 412          people do).
 413          Output directed to console windows needs to be converted (to
 414          GetOEMCP() if the console is using a raster font, or to
 415          GetConsoleOutputCP() if it is using a TrueType font).  Cygwin does
 416          this conversion transparently (see winsup/cygwin/fhandler_console.cc),
 417          converting to GetConsoleOutputCP().  This leads to correct results,
 418          except when SetConsoleOutputCP has been called and a raster font is
 419          in use.  */
 420       sprintf (buf, "CP%u", GetACP ());
 421       codeset = buf;
 422     }
 423 #  endif
 424
 425 # else
 426
 427   /* On old systems which lack it, use setlocale or getenv.  */
 428   const char *locale = NULL;
 429
 430   /* But most old systems don't have a complete set of locales.  Some
 431      (like SunOS 4 or DJGPP) have only the C locale.  Therefore we don't
 432      use setlocale here; it would return "C" when it doesn't support the
 433      locale name the user has set.  */
 434 #  if 0
 435   locale = setlocale (LC_CTYPE, NULL);
 436 #  endif
 437   if (locale == NULL || locale[0] == '\0')
 438     {
 439       locale = getenv ("LC_ALL");
 440       if (locale == NULL || locale[0] == '\0')
 441         {
 442           locale = getenv ("LC_CTYPE");
 443           if (locale == NULL || locale[0] == '\0')
 444             locale = getenv ("LANG");
 445         }
 446     }
 447
 448   /* On some old systems, one used to set locale = "iso8859_1". On others,
 449      you set it to "language_COUNTRY.charset". In any case, we resolve it
 450      through the charset.alias file.  */
 451   codeset = locale;
 452
 453 # endif
 454
 455 #elif defined WINDOWS_NATIVE
 456
 457   static char buf[2 + 10 + 1];
 458
 459   /* The Windows API has a function returning the locale's codepage as a
 460      number: GetACP().
 461      When the output goes to a console window, it needs to be provided in
 462      GetOEMCP() encoding if the console is using a raster font, or in
 463      GetConsoleOutputCP() encoding if it is using a TrueType font.
 464      But in GUI programs and for output sent to files and pipes, GetACP()
 465      encoding is the best bet.  */
 466   sprintf (buf, "CP%u", GetACP ());
 467   codeset = buf;
 468
 469 #elif defined OS2
 470
 471   const char *locale;
 472   static char buf[2 + 10 + 1];
 473   ULONG cp[3];
 474   ULONG cplen;
 475
 476   /* Allow user to override the codeset, as set in the operating system,
 477      with standard language environment variables.  */
 478   locale = getenv ("LC_ALL");
 479   if (locale == NULL || locale[0] == '\0')
 480     {
 481       locale = getenv ("LC_CTYPE");
 482       if (locale == NULL || locale[0] == '\0')
 483         locale = getenv ("LANG");
 484     }
 485   if (locale != NULL && locale[0] != '\0')
 486     {
 487       /* If the locale name contains an encoding after the dot, return it.  */
 488       const char *dot = strchr (locale, '.');
 489
 490       if (dot != NULL)
 491         {
 492           const char *modifier;
 493
 494           dot++;
 495           /* Look for the possible @... trailer and remove it, if any.  */
 496           modifier = strchr (dot, '@');
 497           if (modifier == NULL)
 498             return dot;
 499           if (modifier - dot < sizeof (buf))
 500             {
 501               memcpy (buf, dot, modifier - dot);
 502               buf [modifier - dot] = '\0';
 503               return buf;
 504             }
 505         }
 506
 507       /* Resolve through the charset.alias file.  */
 508       codeset = locale;
 509     }
 510   else
 511     {
 512       /* OS/2 has a function returning the locale's codepage as a number.  */
 513       if (DosQueryCp (sizeof (cp), cp, &cplen))
 514         codeset = "";
 515       else
 516         {
 517           sprintf (buf, "CP%u", cp[0]);
 518           codeset = buf;
 519         }
 520     }
 521
 522 #endif
 523
 524   if (codeset == NULL)
 525     /* The canonical name cannot be determined.  */
 526     codeset = "";
 527
 528   /* Resolve alias. */
 529   for (aliases = get_charset_aliases ();
 530        *aliases != '\0';
 531        aliases += strlen (aliases) + 1, aliases += strlen (aliases) + 1)
 532     if (strcmp (codeset, aliases) == 0
 533         || (aliases[0] == '*' && aliases[1] == '\0'))
 534       {
 535         codeset = aliases + strlen (aliases) + 1;
 536         break;
 537       }
 538
 539   /* Don't return an empty string.  GNU libc and GNU libiconv interpret
 540      the empty string as denoting "the locale's character encoding",
 541      thus GNU libiconv would call this function a second time.  */
 542   if (codeset[0] == '\0')
 543     codeset = "ASCII";
 544
 545 #ifdef DARWIN7
 546   /* Mac OS X sets MB_CUR_MAX to 1 when LC_ALL=C, and "UTF-8"
 547      (the default codeset) does not work when MB_CUR_MAX is 1.  */
 548   if (strcmp (codeset, "UTF-8") == 0 && MB_CUR_MAX <= 1)
 549     codeset = "ASCII";
 550 #endif
 551
 552   return codeset;
 553 }